Visualization of Biomass Pyrolysis and Temperature Imaging in a

Energy & Fuels 2009, 23, 993–1006

993

Visualization of Biomass Pyrolysis and Temperature Imaging in a Heated-Grid Reactor M. J. Prins,*,† J. Lindeń,‡ Z. S. Li,‡ R. J. M. Bastiaans,† J. A. van Oijen,† M. Aldeń,‡ and L. P. H. de Goey† Combustion Technology Group, EindhoVen UniVersity of Technology, Post Office Box 513, 5600 MB EindhoVen, The Netherlands, and DiVision of Combustion Physics, Lund UniVersity, Post Office Box 118, S-221 00 Lund, Sweden ReceiVed June 2, 2008. ReVised Manuscript ReceiVed NoVember 28, 2008

The main advantage of a heated-grid reactor for studying pyrolysis kinetics of solid fuel samples is that high heating rates of up to 1000 K/s can be obtained. However, one of the concerns is whether the temperature distribution over the grid material is uniform and whether the presence of a thermocouple welded to the grid causes any measurement errors. Biomass samples were placed on the heated-grid reactor, and the volatiles, emitted in the biomass pyrolysis process as hot gas plumes, were imaged with an infrared camera with a high framing speed. The temporal resolved infrared images indicate that the pyrolysis process does not take place at the same rate everywhere on the grid. Two-dimensional temperature images of a heated grid made of stainless steel were recorded using the method of laser-induced thermometry with thermographic phosphors. As expected from a heat-transfer model, measured temperatures were found to be significantly higher than temperatures indicated by a thermocouple welded to the bottom of the grid. It was also observed that there is a large temperature gradient between the two electrodes on which the grid is connected. It is shown that replacing a wire mesh by a foil as a grid material may lead to more homogeneous temperature distribution. The paper recommends additional research to demonstrate the suitability of the heated-grid reactor for carrying out accurate measurements.

1. Introduction To study the thermal decomposition of solid fuel materials, such as coal, biomass, and waste, the heated-grid reactor has been used for over 25 years. It generally consists of a wire mesh, which is electrically heated, and connected to a thermocouple for measuring its temperature. This type of reactor facilitates the characterization of solid samples at high heating rates up to 1000 K/s. The main benefit involved here is that the weight loss during heating of the sample (which invariably happens in other methods, such as thermogravimetric analysis) can be minimized. Therefore, the reactivity of the fuel is not altered before it reaches the final temperature, at which decomposition kinetics are studied. Another benefit is that the devolatilization products enter directly into a cold gas phase, so that they are quenched. This minimizes secondary reactions, so that the primary pyrolysis gases can be determined. Initial experiments1-3 have been carried out under so-called “zero hold time” conditions, where the grid was heated to the final temperature and, immediately, the power was turned off. In most experiments, samples are kept at the final temperature, so that kinetic data * To whom correspondence should be addressed. E-mail: mark.prins@ shell.com. † Eindhoven University of Technology. ‡ Lund University. (1) Suuberg, E. M.; Peters, W. A.; Howard, J. B. SeVenteenth International Symposium on Combustion; The Combustion Institute: Pittsburgh, PA, 1979; pp 117-130. (2) Desypris, J.; Murdoch, P.; Williams, A. Fuel 1982, 61, 807–816. (3) Unger, P. E.; Suuberg, E. M. Fuel 1984, 63, 606–611.

can be obtained.4 For kinetic studies, it may be important (especially when pyrolysis is incomplete) to minimize also the cooling time. The wire mesh will typically be cooled by radiation and natural convection at a rate of several hundred degrees per second, but cooling rates can be increased by forced convection.5 More recent research describing the use of a heatedgrid reactor is available.6,7 Despite its advantages, results obtained from a heated-grid reactor should be handled with care. First of all, it must always be verified whether the particle size and, hence, the Biot number, is small enough to ensure that the particles on the grid can follow the large temperature gradient imposed. Typically, a fine powder with a particle size of several hundreds of micrometers up to a few millimeters must be used. As a consequence for the reactor design, the aperture between the wires of the mesh cannot be too large. Second, the thermal load on the grid is restricted. If too much sample material were placed on the wire mesh, the amount of electric energy dissipated into heat will be insufficient for rapid heating. Finally, the question is whether all of the particles of the solid sample are heated at the same rate, i.e., whether heating takes place in a uniform way. This requires that the particles are evenly distributed over the grid, so that the thickness of the sample layer is the same everywhere. Another requirement is that the temperature distribution over (4) Niksa, S. J.; Russel, W. B.; Saville, D. A. Nineteenth International Symposium on Combustion; The Combustion Institute: Pittsburgh, PA, 1982; pp 1159-1167. (5) Freihaut, J. D.; Proscia, W. M. Energy Fuels 1989, 3, 625–635. (6) Mu¨hlen, H.-J.; Sowa, F. Fuel 1995, 74 (11), 1551–1554. (7) De Jong, W. Nitrogen compounds in pressurised fluidised bed gasification of biomass and fossil fuels. Ph.D. Thesis, Delft University of Technology, Delft, The Netherlands, 2005.

10.1021/ef800419w CCC: $40.75  2009 American Chemical Society Published on Web 02/02/2009

994 Energy & Fuels, Vol. 23, 2009

the grid material is uniform. During calibration measurements, in which a sample of solid bismuth was melted, slight nonisothermalities have been observed.5 The bismuth melted at the edges of the grid before it melted at the center, where the thermocouple was located. The presence of the thermocouple, with a radius of the same magnitude as the wires of the grid, may disturb the thermal field. A model developed by Guo8 shows that, because of heat loss via the thermocouple, the indicated temperature may be significantly lower than the average grid temperature. Using an optical pyrometer, Guo measured grid temperatures that were 150-200 K higher than the thermocouple temperature in the range of 800-1600 K. The primary aim of this paper is a qualitative assessment of the pyrolysis rate of biomass, which should be the same everywhere on the heated grid. Some insight can be obtained by measuring the temperature distribution of the heated grid, in the absence of a solid sample on the grid, using methods that are described later. To verify whether the biomass is pyrolyzed in a uniform way, i.e., independent of its horizontal and vertical position on the grid, experiments are also carried out in the presence of biomass. Because infrared active gases, e.g., carbon dioxide, are produced during the biomass decomposition procedure, infrared emission can be used to visualize qualitatively the distribution of hot volatiles. Temporal resolved infrared images of the hot volatile gas plumes were recorded during the rapid gas pyrolysis process. Another aim of this work is to study the 2D temperature distribution of the heated grid, investigate the relation between the temperature of the wire mesh and temperature indicated by the thermocouple, and verify the model of Guo. The method for measuring temperature involves laser-induced emission from thermographic phosphors, which enable remote temperature diagnostics to be performed with high sensitivity and accuracy. The technique is superior to methods based on thermocouples and pyrometry. An optical pyrometer, focusing the radiation from the hot platinum grid onto a screen, will only work at high temperatures and requires focusing time (during which time the grid may break down because of thermal fatigue). The method applied in this research works quickly and is valid over a wide temperature range. This paper comprises principles of the applied spectroscopic methods, a description of the experimental setup, and discussion of results, including the validity of specific heat-transfer models described below. 2. Heat-Transfer Theory The systematic measurement error in the temperature of the heated grid, because of the presence of the thermocouple leading to disturbances of the temperature field, is examined in detail in this paper. Other sources of systematic errors,9 such as intrinsic errors in the circuit, extraneous signals or influences, and errors in signal processing, are not considered. Another important factor to avoid measurement errors is to make sure that the thermocouple wires are firmly welded onto the object, of which the temperature is to be measured. To obtain an idea of how significant the systematic error can be, Keltner and Beck10 derived a model that calculates the temperature difference between a thermocouple and a thick wall, against which the thermocouple is pressed. The model is based (8) Guo, J. Pyrolysis of wood powder and gasification of wood-derived char. Ph.D. Thesis, Eindhoven University of Technology, Eindhoven, The Netherlands, 2004. (9) Bentley, R. E. Handbook of Temperature; Springer: Singapore, 1998. (10) Keltner, N. R.; Beck, J. V. J. Heat Transfer 1983, 105, 312–318.

Prins et al.

on the unsteady surface element method.11 The steady-state solution of the model calculates the temperature measured by the thermocouple wires (Ts) as function of the actual wall temperature (Twall) Ts )

Twall exp(-4√Bi) 1 π 1 + 2K√Bi + B 4

(

(1)

)

with K)

kTc kwall

B)

hrTc kwall

Bi )

hTcrTc 2kTc

In the equations, k is the heat conduction coefficient of the metals, h is the contact heat-transfer coefficient between the wall and the thermocouple, while hTc is the heat-transfer coefficient of the thermocouple to the ambient, and rTc is the radius of the thermocouple. The grid cannot be compared to a thick wall, because its height is in the order of 100 µm, whereas the diameter of the thermocouple is of the same order. Guo8 developed a model to predict the temperature error in this situation; the derivation of this model can be found in Appendix A. It is essentially an energy balance for the weld, which is considered to have an isothermal temperature Ts. The energy losses from this volume to the thermocouple, the environment (by convection and radiation), and the rest of the grid have to be compensated by dissipation of electric energy. The model is valid for negligible thermal resistance of the interface (i.e., B f ∞). It assumes that the temperature of the thermocouple only varies in the z direction and the temperature of the grid only varies in the r direction. The steady-state result is given in eq A38 for the situation where one wire is attached to the grid. For a thermocouple, which consists of two wires welded together onto the grid, the radius of the weld is approximately 1.4 times the radius of a thermocouple wire12 rs ) √2rTc

(2)

with rTc being the radius of a single thermocouple wire. Assuming that both wires have the same radius and (more or less) the same heat-transfer properties, eq A38 changes to Ts ) Ta

(

) ( (

2hTckTc - hgrid + T∞ 2hgrid + rTc hgrid +

8hgridδkgrid K1(r′s) K0(r′s) r2 s

8hgridδkgrid K1(r′s) K0(r′s) r2

2hTckTc + rTc

s

)

)

(3)

with r′s )

rs2

2hgrid δkgrid

From these expressions, it is clear that the measured temperature depends upon the thermal conductivity of the grid as well as the thermocouple, their heat-transfer coefficients to the gaseous environment, the thickness of the grid (which equals the diameter of the wires), the grid temperature, and the ambient temperature. (11) Keltner, N. R.; Beck, J. V. J. Heat Transfer 1981, 103, 759–764. (12) Hill, J. M.; Dewynne, J. N. Heat Conduction; Blackwell Scientific Publications: London, U.K., 1987.

Heated-Grid Reactor for Pyrolysis Kinetic Studies

Energy & Fuels, Vol. 23, 2009 995

Figure 1. Temperature measurement error predicted by models of Keltner and Beck and Guo. Parameters: kTc ) 70 W m-1 K-1; hTc ) 140 W m-2 K-1; kgrid ) 70 W m-1 K-1; δ ) 100 µm; Tgrid ) 1000 K; Tambient ) 300 K.

Figure 1 compares the relative measurement error of the model of Keltner and Beck (for a thermocouple attached to an infinite plate) and the model of Guo (for a thermocouple attached to an infinite plate but of finite thickness; the plate loses energy to the environment and is heated by electric energy) as a function of the thermocouple radius. In both cases, perfect contact between the thermocouple and object is assumed. If the radius of the thermocouple is increased, the temperature measured by the thermocouple will decrease, because of the heat loss through the thermocouple, leading to higher temperature measurement error. The errors predicted are rather similar for a typical value of the heat-transfer coefficient from the plate to ambient of 140 W m-2 K-1. If this heat-transfer coefficient is much smaller, this also means that much less electric energy is needed to maintain the temperature of the plate. As a consequence, the heat loss via the thermocouple has a much bigger influence and the measurement error increases. On the other hand, if heat transfer from the grid to ambient is extremely high, it may happen that the temperature in the weld becomes higher than the temperature of the rest of the grid. However, this is a rather theoretical situation. 3. Spectroscopic Methods 3.1. Visualization of Volatile Gas Plumes. Real-time visualization of the generated volatile gas flows during the pyrolysis is attractive to understand the reaction kinetics of the process. However, this is not an easy task because the gas components are invisible, with regard to the lack of electronic transitions probed for these molecular species. Carbon dioxide and water are among the major components in the volatiles. Although invisible in the ultraviolet/visible spectral range, they are infrared-active. The infrared radiation from these hot volatile carbon dioxide and water vapor can be used for visualization of fresh volatile plumes from the heat-grid reactor. Because of the relatively high infrared radiation probability of CO2 at 4.5 µm from its unsymmetric stretching vibration, the major infrared radiation is expected to be located around this wavelength.13,14 3.2. Temperature Measurements Using Laser-Induced Phosphorescence. The use of thermographic phosphor materials is a recent technique, which has been developed for remote (13) Clausen, S.; Bak, J. J. Quant. Spectrosc. Radiat. Transfer 1999, 61, 131–141. (14) Bak, J.; Clausen, S. Meas. Sci. Technol. 2002, 13, 150–156.

temperature measurements in various applications.15 These include static as well as moving surfaces, such as rotor engines16,17 and turbine engines.18 Some attempts have been made to apply the method to combustion,19 e.g., for flame spread scenarios.20 For thermal decomposition, i.e., pyrolysis, laserinduced phosphorescence was pioneered by Omrane et al.21 and Svenson et al.,22 who studied the surface temperature of decomposing materials. The technique is based on exciting a thermographic phosphor material by an appropriate light source. The phosphor becomes highly fluorescent or phosphorescent, and typically, the emission is in the visible region, with a lifetime in the order of 1 ms. These phosphors used for temperature measurements consist of a ceramic host material and a doping agent from which light is emitted. A large number of different phosphors are known today, covering a wide range from cryogenic temperatures up to 1600 °C or higher. For every application, a specific phosphor can be selected with a typical accuracy of 1% or better, depending upon the conditions. These temperature measurements can essentially be performed in two ways: (1) Temporally resolved measurements; these are based on the principle that the decay of the phosphorescence signal is temperature-dependent. At higher temperatures, the lifetime becomes shorter. (2) Spectrally resolved measurements; these are based on monitoring the emission versus wavelength. Thermographic phosphors emit phosphorescence signals at different wavelengths; some of these wavelengths are more temperature-sensitive than others. Figure 2 illustrates how the (15) Allison, S. W.; Gillies, G. T. ReV. Sci. Instrum. 1997, 68 (7), 2615– 2650. (16) Allison, S. W.; Cates, R. M.; Noel, W. B.; Gillies, G. T. IEEE Trans. Meas. 1988, 37 (4), 637–641. (17) Omrane, A.; Juhlin, G.; Aldeń, M.; Josefsson, G.; Engstro¨m, J.; Benham, T. The Society of Automotive Engineers World Congress, Session: CI and SI Power Cylinder Systems (Parts 1 and 2), Detroit, MI, 2004; paper 2004-01-0609. (18) Alaruri, S.; Bonsett, T.; Brewington, A.; McPheeters, E.; Wilson, M. Opt. Lasers Eng. 1999, 31, 345–351. (19) van der Wal, R. L.; Householder, P. A.; Wright, T. W. Appl. Spectrosc. 1999, 53 (10), 1251–1258. (20) Omrane, A.; Ossler, F.; Aldeń, M. Exp. Therm. Fluid Sci. 2004, 28, 669–676. (21) Omrane, A.; Ossler, F.; Aldeń, M.; Svenson, J.; Pettersson, J. B. C. Fire Mater. 2005, 29, 39–51. (22) Svenson, J.; Pettersson, J. B. C.; Omrane, A.; Ossler, F.; Aldeń, M.; Bellais, M.; Lilliedahl, T.; Sjo¨stro¨m, K. Proceedings of the science in thermal and chemical biomass conversion conference, Victoria, British Columbia, Canada, 2004.


Prins et al.

Figure 2. YAG:Dy spectra at different temperatures. When the temperature increases, the ratio from 455 to 493 nm increases. This figure was reproduced with permission from Omrane.23

Figure 3. Schematic picture of the heated-grid reactor.

temperature can be derived from taking the ratio of the emission at two different wavelengths, in this case, 455 and 493 nm. The precision of the first method, i.e., the lifetime technique, is 1-5 K, whereas for the second method, i.e., the ratio technique, it is 5-10 K. Both methods are able to give 2D or even 3D images and avoid interferences from laser scattering and broadband (nanosecond) fluorescence. The second method has the advantage of less interference with blackbody emission and is therefore applied for this research, because its accuracy is sufficiently high. Another benefit of the ratio technique is that it is insensitive to fluctuations in the laser profile over the surface that is studied, because two images are overlaid, so that any fluctuations are always corrected. 4. Experimental Section 4.1. Heated-Grid Reactor. Figure 3 shows a picture of the heated grid reactor. This consists of an optically accessible reactor cell, which contains a heated grid with dimensions of 1.6 × 1 cm. Optical windows made of fused silica are located on the top and both sides of the reactor, so that the process can be investigated with an intensified CCD camera. The grid can be heated quickly by electrodes on both sides of the grid. To power the grid, a control box was designed and built in-house, containing three main components: a Delta power source, a mini PC, and a Keithley data acquisition module. The Delta power source is a type S 6-40, with a voltage range of 0-6 V and a current range of 0-40 A. To read the thermocouple that is fixed to the grid, a Keithley module is used. The KUSB_3108 has cold junction compensation (CJC) and a reading rate of 50 kHz. Finally, a mini-PC is built into the box of the power source. The task of the mini-PC, which uses Labview software, is to control the Delta power source; the control loop (23) Omrane, A. Thermometry using laser-induced emission from thermographic phosphors: Development and applications in combustion. Ph.D. Thesis, Lund University, Sweden, 2005.

needs to be very quick to ensure that the grid reaches the desired temperature quickly enough without excessive overshoot. The electric energy required to maintain a certain grid temperature (below 700 °C) was found to increase linearly with temperature, as shown in Figure 4, which agrees with observations of Freihaut and Proscia.5 This is as expected when heat loss mechanisms have insufficient time to start their exchange. The grid consists of AISI 316 L stainless steel. Three different mesh sizes have been used, which are referred to as extra fine, fine, and coarse (see Table 1 for the respective wire diameters and apertures). For some experiments, the grid was coated with a thin layer (3 ( 1 µm) of ceramic AlCrN coating (available from Oerlikon Balzers Benelux NV). The purpose of this coating is to avoid direct contact of the solid sample with the Cr- and Nicontaining metal, so that possible catalytic effects are avoided. The temperature of the grid is measured by a thermocouple, existing of two un-insulated wires of a different material, welded to the grid. In the temperature range of 400-1000 K, which is of interest for pyrolysis of biomass and coal, a K-type thermocouple is frequently applied. This type has a steady, linear relation between the thermal electromotive forces and the temperature.22 4.2. Chemiluminescence Experiments. These experiments were carried out in the presence of biomass on the heated grid. For this purpose, two side windows of the reactor cell were replaced with sapphire plates to enable the transmission of infrared light from hot volatile gases. An infrared camera used in this experiment was a 256 × 256 pixel InSb camera (Santa Barbara Focal Plane, SBF LP134), with a spectral response cover 1.2-5.5 µm. When the biomass was pyrolyzed, the emitted hot volatile gas plumes were recorded horizontally by the infrared camera through the sapphire window at a frame rate of 50 Hz (maximum is 200 Hz). To avoid the interference from the strong blackbody radiation from the hot heating wires, a black screen was used to block the heating wire in the vision of the infrared camera. All experiments were performed with pinewood in pulverized form and sieved to a particle diameter range of 0.1-0.2 mm. The amount of wood placed on the grid varied from 3 to 6 mg. After carefully weighing the biomass and placing it on the grid (an extra fine grid was used), the reaction chamber was flushed with nitrogen to create an inert atmosphere. This was performed carefully to avoid blowing the biomass from the grid. After each experiment, the grid is replaced, because it is very difficult to remove residual char. In the experiments, the temperatures were varied from 500 to 700 °C, the heating rates were varied from 300 to 600 °C/s, and the holding time was varied from 7 to 20 s. 4.3. Thermometry Experiments. The metal wire mesh (uncoated stainless steel, coarse; coated stainless steel, coarse; and uncoated stainless steel, fine) of 1.6 × 1 cm is impregnated with a thin layer of YAG:Dy phosphore (Y3Al5O12:Dy; yttrium alumini-



Figure 4. Electricity energy required to maintain the grid temperature; data points calculated from measured current and voltage to maintain the grid temperature. Table 1. Dimensions of Grids Used in Experiments mesh wire diameter (mm) aperture (mm) mesh width (mm) N (number of wires/m)

extra fine

fine

coarse

250 0.04 0.062 0.102 9804

150 0.065 0.104 0.169 5917

60 0.16 0.263 0.423 2364

Subsequently, the set point for the grid temperature, as measured by the thermocouple, was increased in steps of 50 °C up to a final temperature of 800 °C. At each set temperature, the time was allowed to stabilize and, subsequently, thermometry measurements were carried with the aim of determining the temperature distribution of the grid. An alternative thermometry method was also applied. At the high end of the temperature range, the grid emits radiation in the visible range, which was recorded with a standard photo camera. By calibrating with objects at thermal equilibrium, it was possible to determine a temperature scale for the recorded images. This method was used to study the effect of replacing the wire mesh with a metal foil, to check whether the shape of the grid has any influence on its temperature distribution.

5. Results

Figure 5. Experimental setup for thermometry measurements.

umoxid dysprosium) and placed between two electrodes. The experimental setup comprising the laser, optical filter, detector, stereroscope, and reactor cell is shown in Figure 5. For thermometry measurements, the top window of the reactor cell is used. Laser light from a NdYAG laser (manufactured by Quantel) is transformed by frequency doubling and tripling to a wavelength of 355 nm. Pulse energy was approximately 10 mJ/pulse. The phosphoresence signal from the thermographic phosphor on the grid was detected with a Princeton Instruments ICCD camera, with a frame rate of 10 frames per second. A gate delay of 120 ns was applied to filter away the laser. The camera was used in connection with a stereoscope and two optical filters at 455 and 493 nm. By overlaying the two pictures obtained and taking the ratio of the signals, a 2D temperature image was determined. The resolution of the ICCD is equivalent to 5 pixels per the smallest (square) element of the finest grid; the other grids had more pixels per element. White image correction was applied to ensure that measured temperatures are not influenced by the spectroscope or other optic distortions. Prior to any measurements, the method was calibrated by measuring on a surface target, of several square centimeter, placed in an oven at thermal equilibrium. The method may also be verified with measurements of a well-calibrated IR camera.

5.1. Results of Chemiluminescence Experiments. The infrared natural emission from the volatiles produced in biomass pyrolysis was visualized, for different grid temperatures. Parts a-f of Figure 6 and parts a-f of Figure 7 show the results for a temperature set point of 500 and 800 °C, respectively. The first picture, shown in Figure 6a and, respectively, Figure 7a, was taken immediately after the current through the grid drops (because the desired thermocouple temperature was reached), and subsequent pictures shown were recorded with 0.1 s intervals. These pictures show that the biomass particles on the heated grid are pyrolyzed faster at the higher temperature and that more volatiles are formed at the higher temperature. This is in agreement with literature in this field, e.g., ref 24; devolatilization of biomass is more rapid and more complete at higher temperatures. Although it is difficult to see properly, the chemiluminescence experiments show that the volatiles are not produced as an evenly distributed thermal plume. It is clear that the first volatiles are formed on the left side, and only later, more volatiles are formed on the right side of the grid. This effect is most (24) Antal, M. J. Biomass pyrolysis: A review of the literature. Part IIsLignocellulose pyrolysis. In AdVances in Solar Energy; Boer, K. W., Duffie, J. A., Eds.; American Solar Energy Society: Boulder, CO, 1985; Vol. 2, pp 175-255.


Prins et al.

Figure 6. (a-f) Two-dimensional images of natural emission from volatiles produced by pyrolysis of biomass particles on the heated grid. Grid temperature ) 500 °C; amount of biomass ) 2 mg on a 1 cm2 grid; pictures are taken with 0.1 s intervals (note: the two half circles on the pictures are reflections from the reactor wall).

Figure 7. (a-f) Two-dimensional images of natural emission from volatiles produced by pyrolysis of biomass particles on the heated grid. Grid temperature ) 800 °C; amount of biomass ) 4 mg on a 1 cm2 grid; pictures are taken with 0.1 s intervals (note: the two half circles on the pictures are reflections from the reactor wall; these are stronger compared to Figure 6 because of the higher temperature used in this measurement).

noticeable at the lower temperature in Figure 6c. At higher grid temperatures, the formation of volatiles as seen by the chemiluminescence in Figure 7 takes place more evenly. The observed phenomena, which were unexpected, may be explained by the presence of temperature gradients on the grid. This would mean that that the temperature on the left side is higher than on the right side of the picture, especially when the grid temperature is relatively low. To confirm whether this is

indeed the case, thermometry measurements were carried out, which are described in the next section. 5.2. Results of Thermometry Experiments. For all of the grids used in experiments (coarse grid, coarse grid with AlCrN coating, and fine grid), the actual temperatures as determined by laser-induced thermometry with thermographic phosphors were higher than the temperatures indicated by the thermocouple. As explained in the section on heat-transfer theory, this



Figure 8. (a-c) Two-dimensional image of the measured temperature on the grid for thermocouple temperatures of (a) 400 °C, (b) 500 °C, and (c) 600 °C (note: the shape is oval because the edges of the grid have been omitted).

result was expected. Furthermore, parts a-c of Figure 8 confirm that the temperature on the grid has a gradient from one electrode to the other electrode. Again, this was observed for all of the different grids. The temperature distribution seemed to become more equal at higher temperatures. Nevertheless, the temperature on one side of the grid (the left-hand side in parts a-c of Figure 8) remained higher than on the other side. The average grid temperature measured was determined, as well as the average error based on a 95% interval. This average error is caused by measurement uncertainty (which is only 1-5 K) and temperature variation on the heated grid (which gives a

much larger contribution). Parts a-c of Figure 9 compare these measured temperatures with temperatures predicted by the model of Guo for two wires, i.e., eq 3. For the latter case, which is the most realistic, Table 2 shows that the relative measurement error predicted may range from 10-15% at 300 °C to 16-25% at 800 °C (the lower numbers are for the coarse grid, and the higher numbers are for the fine grid). For all grids, there is a very large measurement error at low temperatures, which happens because the thermocouple (in the middle of the oval pictures in Figure 5) is still far away from the hot zone in the left of the picture. At higher temperatures, the measurement error becomes smaller


Figure 9. Continued.

Prins et al.



Figure 9. (a-d) Comparison of grid temperatures, experimentally measured by laser-induced thermometry and predicted by eq 3, as a function of the measured thermocouple temperature, for (a) uncoated grid, coarse; (b) coated grid, coarse; (c) uncoated grid, fine; and (d) model predictions. Table 2. Relative Temperature Measurement Error Predicted from the Model by Guo for the Heated Grid of AISI 316L Stainless Steel for one wire

for two wires

Coarse Grid thermocouple temperature ) 300 °C thermocouple temperature ) 800 °C

0.06 0.10

0.10 0.16

Fine Grid thermocouple temperature ) 300 °C thermocouple temperature ) 800 °C

0.10 0.17

0.15 0.25

because of a more homogeneous temperature distribution; for the uncoated grids in parts a and c of Figure 9, it becomes even less than the error predicted by the model. For the coated grid in Figure 9b, the measurement error at higher temperatures appears to be more consistent with the model. However, it must be realized that the model assumes isothermal conditions, except for the area to which the thermocouple wires are welded. The situation is very different in reality, which invalidates the model. These results could be explained in several different ways: (1) The hypothesis was formulated that not only heat but also a current may leak away through the thermocouple. This hypothesis was tested by performing a few experiments without thermocouple wires welded to the grid. However, this had little effect on the temperature distribution, which invalidated the hypothesis. (2) Second, it is possible that the emissivity coefficient of the grid rises substantially at elevated temperatures, so that radiation becomes much stronger. The benefit of increased radiation is that more electric energy is dissipated in the grid, so that the heat lost via the thermocouple plays a relatively smaller role. However, this explanation is not very likely, because most of the heat lost from the grid is a convective heat loss, even if the grid were a blackbody. Our calculations, in line with those of Guo, show that the radiative heat-transfer coefficient is at least 5 times smaller than the convective heattransfer coefficient at temperatures around 1000 °C. Furthermore, it must be remembered that there is a layer of white thermographic phosphor present on the grid, which actually suppresses radiation (although radiative heat transfer between the wires of the grid can take place). (3) Probably, the temperature differences stem from the relatively poor thermal

conductivity of the stainless-steel grid; with increasing temperatures, the temperature distribution becomes more homogeneous as the grid conducts heat better. This may explain the difference between the uncoated grid (parts a and c of Figure 9) and coated grid (Figure 9b); changes in thermal conductivity, when a ceramic coating is applied, could give rise to a change in temperature homogeneity of the grid. For the stainless-steel grids, the thermal conductivity may increase with temperature, but this effect was not taken into account in the model. For the grids that are coated with the AlCrN layer, the thermal conductivity does not seem to improve with temperature. In fact, thermal conductivity is one of the most important parameters in the model. A better heat-conducting grid should make the grid more homogeneous in temperature and also decrease the measurement error. For platinum grids, which have a much higher conductivity (around 58 W m-1 K-1) than stainless steel, the measurement errors predicted by the model are only 2.0-2.2% compared to 6-17% (for the dimensions of the coarse grid; error increases with temperature). Alternatively, stainless steel could be used as a foil rather than as a wired mesh; the disadvantage of a mesh is that, even when thicker wires are used, because of construction reasons also, bigger apertures must be applied. Therefore, an experiment was carried out using a nickel/chromium (Ni80/Cr20) foil with a thickness of 0.1 mm as grid material. It was carried out at a thermocouple temperature of 725 °C; the 2D temperature distribution was measured and shown in Figure 10. Note that, because the temperature measurement is based on radiation, temperatures shown below 600 °C have no physical meaning. In Figure 10a, it can be observed that a more homogeneous temperature profile can indeed be obtained. Furthermore, because of the higher thermal mass when compared to a wired mesh, the effect of a thermal leak via the thermocouple does not appear to have an effect. Heat losses from the sides of the foil do have an influence, because the temperature is highest in the middle. Finally, Figure 10b shows what happens when the foil is not fixed very tightly to the electrodes. In this case, the bottom right of the foil is loose (because a screw was not tightly fixed) and it stays relatively cold. In the future, more research is recommended using a foil, also at lower temperatures using


Prins et al.

Figure 10. (a and b) Comparison of grid temperatures, experimentally measured by capturing visible emission with a standard camera, at a thermocouple temperature of 725 °C: (a) good contact between the grid and clamp holders/electrodes and (b) insufficient contact between the grid and clamp holders/electrodes.

thermographic phosphors or IR camera, with the final aim to demonstrate homogeneous devolatilization of solid fuel on the grid. 6. Conclusions and Recommendations The natural emission from volatiles produced in biomass pyrolysis in a heated-grid reactor was visualized; this process does not take place at the same rate everywhere on the grid. It was also observed that there is a significant temperature gradient between the two electrodes on which the grid is connected. Two-dimensional temperature images of the temperature on a heated grid were successfully recorded using the method of

laser-induced thermometry with thermographic phosphors. It was found that the measured temperatures were significantly higher than temperatures indicated by a thermocouple welded to the bottom of the grid. This was in line with expectations from heattransfer models that take the systematic measurement error, because of the presence of the thermocouple leading to disturbances of the temperature field, into account. To improve both the measurement error and the uniformity of the temperature, it was suggested to use a material with high heat conductivity and/or in the form of a foil rather than a wire mesh. Indeed, a more homogeneous temperature profile was observed from an experiment using a Ni/Cr foil as grid material.



Heat Loss from the Weld to the Rest of the Heated Surface. This term may be derived by setting up a heat balance, similar to eq A1, for a shell of the horizontal surface, between r and r + dr q1,top + q1,bottom ) qe + q2,r - q2,r+dr

(A3)

where q1,bottom is the heat loss from the bottom of the heated wire to the gaseous environment. On both the top and the bottom of the grid, the heat flow to the environment is supposed to be equal to each other, assuming the ambient gas has the same conditions on both sides of the grid. This leads to q1,top ) q1,bottom ) 2πrdrhgrid(T - Ta)

(A4)

The heat produced by dissipation of electric energy in the shell of the surface can be expressed as Figure A1. Geometry for a thermocouple attached on the surface of the grid.

Additional research, both in the absence and presence of biomass on the heated grid, is recommended to confirm the suitability of the heated-grid reactor for pyrolysis experiments. Acknowledgment. Bo Li is thanked for his kind assistance with the experiments, and Dr. Alexei Sepman is thanked for providing Figure 10. We are also grateful for the financial support of CECOST (Centre for Combustion Science and Technology) and the European Union in the framework of the large-scale facility.

Appendix A: Model for Temperature Measurement Accuracy of Heated Wire Mesh A1. Heat Balance for Thermocouple Wire Attached to a Horizontal Surface. Figure A1 shows a horizontal surface to which a vertical thermocouple wire of radius rTc is welded. An initially uniform temperature distribution, in the absence of a thermocouple, is disturbed by the application of the thermocouple to the grid acting as a heat sink. To derive the magnitude of the heat flow through this heat sink, the spot where the thermocouple wire is welded to the grid is examined separately. This welded spot is assumed to be an isothermal disk with temperature Ts and radius rTc and has the same height and material as the grid. The horizontal surface is assumed to be infinite, so that it extends along the r coordinate from rTc to infinity; likewise the thermocouple has infinite length in the z direction. The essence of the model developed by Guo8 is a heat balance for the weld, the shaded area shown in Figure A1 q1,top + q3 ) qe,weld - q2,rTc

(A1)

where q1,top is the heat loss from the top of the weld to the gaseous environment, q2,rTc is the heat conducted from the weld to the rest of the heated surface (which will usually have a negative sign because the temperature in the weld is expected to be lower than the temperature in the rest of the heated surface), q3 is the heat conduction from the weld to the thermocouple, and qe,weld is the electric energy dissipation in the weld. The individual terms of this equation will now be derived. Heat Loss from the Weld to the Gaseous EnVironment. This term can be expressed by q1,top ) πrTc2hgrid(Ts - Ta)

(A2)

where hgrid is the combined radiative and convective heat-transfer coefficient from the grid to the gaseous environment, Ts is the uniform temperature of the weld, thus, the temperature actually measured by the thermocouple, and Ta is the temperature of the ambient gas.

qe ) 2πrdrδSe

(A5)

The term Se is the electrical heat input per unit of volume. The net heat flow is described by

(

)

dTr dTr+dr dr dr 2 dT 1 dT dr dr + r dr dr2

q2,r - q2,r+dr ) -2πrδkgrid

(

) 2πrδkgrid

)

(A6)

where kgrid is the effective thermal conductivity of the grid material, which is assumed to be constant, and a Taylor series expansion was applied as in Hill and Dewynne.12 Filling the terms of eqs A4-A6 into eq A3 leads to 4πrdrhgrid(T - Ta) ) 2πrdrδSe + 2πrδkgrid

(

)

d2T 1 dT dr dr + r dr dr2 (A7)

which can be rewritten to r2

2hgrid Se d2T dT + r - r2 (T - Ta) + r2 )0 2 dr δk k dr grid grid

(A8)

This differential equation describes the temperature distribution within the horizontal surface. The following boundary conditions apply T ) Ts at r ) rTc

(A9)

T f T∞ at r f ∞

(A10)

To solve eq A8, two new variables are introduced T ′ ) T - Ta

(A11)

2hgrid δkgrid

(A12)

r′2 ) r2 which leads to r′2

Seδ d2T ′ dT ′ - r′2T ′ ) -r′2 + r′ 2 dr′ 2hgrid dr′

(A13)

with the boundary conditions T ′ ) Ts - Ta at r′ ) r′Tc )

rTc2

T ′ f T∞ - Ta at r′ f ∞

2hgrid δkgrid

(A14)

(A15)

The equation can be recognized as a form of Bessel’s zero-order equation.25 The general solution of this function, related to this problem, can be written as


Prins et al.

T ′(r′) ) C1I0(r′) + C2K0(r′) +

Seδ 2hgrid

I0 and K0, are the modified Bessel functions of the first and second kind, respectively. I0 increases to infinity with an increasing r′, while K0 is gradually decreasing to zero. Therefore, I0 has to be discarded, and C1 becomes zero. Subsequently, filling in the boundary condition of eq A15 leads to T ′(∞) ) C2K0(∞) +

Seδ Seδ ) 2hgrid 2hgrid

dq3 ) -2πrTchTc(TTc - Ta) dz

(A16)

(A17)

in which Ta is the temperature of the ambient gas. Introducing the Fourier law into the above equation gives the following differential equation: -πrTc2kTc

(

( ) dz2

(A18)

)

so that eq A16 becomes K0(r′) (T ′(r′Tc) - T ′(∞)) + T ′(∞) K0(r′Tc)

q2,rTc ) -2πrTcδkgrid

|

dT r ) rTc dr

( ) |

2hgrid dT ′ r ) rTc δkgrid dr′

(A21) (A22)

To evaluate this expression, the recurrence relation K0′(x) ) -K1(x) is used q2,rTc ) 2πrTcδkgrid

( )

q2,rTc ) 2πrTcδkgrid

2hgrid K1(r′Tc) (T ′(r′Tc) - T ′(∞)) δkgrid K0(r′Tc) (A23)

( )

2hgrid K1(r′Tc) (T - T∞) δkgrid K0(r′Tc) s

( )|

(25) Michalski, L.; Eckersdorf, K.; Kucharski, J. Temperature Measurement, 2nd ed.; Wiley: New York, 2000.

(A28)

2hTc kTcrTc

The following boundary conditions apply: at z equals zero, the temperature is equal to the temperature of the weld (Ts), and if the thermocouple wire is infinitely long, the temperature at the end of the wire will be equal to the temperature of the ambient gas T ′Tc ) T ′s ) Ts - Ta at z ) 0

(A29)

T ′Tc f 0 at z f ∞

(A30)

The most general solution can be derived as T ′Tc(z) ) C1 exp(-mz) + C2 exp(mz)

(A31)

According to the second boundary condition, the relative temperature tends to go to zero if z is increased; thus, this cancels out the second term, leaving only the first term. When the first boundary condition is also applied, the solution becomes T ′Tc(z) ) T ′s exp(-mz)

(A32)

If this solution is inserted into eq A25, the heat flow rate (q3) can finally be derived q3 ) -πrTc2kTc

( )

dTTc ) πrTc2kTc dz

(A24)

dTTc z)0 (A25) dz In this equation, kTc is the thermal conductivity of the thermocouple wire. An expression needs to be derived for the distribution of TTc, the temperature of the thermocouple, in the z direction. If the distance z is increased, the temperature drops because of the heat loss from the thermocouple wire to the environment. The rate of heat loss is a function of the surface area of the thermocouple, depending upon the radius and the length, and the heat-transfer coefficient between the thermocouple and the environment (hTc) q3 ) -πrTc2kTc

- m2T ′Tc ) 0

2hTc kTcrTc

T ′s ) πrTc2kTc

Heat Loss from the Weld to the Thermocouple. This heat conduction term may be derived from a cooling fin analysis. Let z represent the length coordinate of the thermocouple wire. The heat flow rate is described by

(A27)

where

(A20)

Now that the temperature distribution of the heated surface is known, q2,rTc, i.e., the heat conducted from the weld to the rest of the heated surface, may finally be calculated q2,rTc ) -2πrTcδkgrid

) -2πrTchTc(TTc - Ta)

T ′Tc(z) ) TTc(z) - Ta and m2 )

Seδ 1 1 (T ′(r′Tc) - T ′(∞)) T ′(r′Tc) ) K0(r′Tc) 2hgrid K0(r′Tc) (A19)

T ′(r′) )

dz2

d2T ′Tc

Combining the above equations gives the constant C2 C2 )

( ) d2TTc

Introducing a relative temperature (T′Tc) simplifies the differential equation to

The boundary condition of eq A14 gives Seδ T ′(r′Tc) ) C2K0(r′Tc) + 2hgrid

(A26)

2hTc (T - Ta) (A33) kTcrTc s

Dissipation of Electric Energy to Heat in the Weld. The electrical heat production in the volume of the weld is calculated as follows: qe,weld ) AsδSe

(A34)

Because the temperature of the heated surface for r f ∞ equals T∞ - Ta )

Seδ 2hgrid

It is becomes possible to express Se and substitute it into eq A34 qe,weld ) AsδSe ) πrTc22hgrid(T∞ - Ta)

(A35)

Now, having all of the terms of the heat balance of this problem derived, these terms can be combined in the heat balance (eq A1) (26) See http://www.lenntech.com/Stainless-steel-316L.htm (accessed on Oct 30, 2007). (27) Ho, C. Y.; Powell, R. W.; Liley, P. E. J. Phys. Chem. Ref. Data 1972, 1 (2), 279–422. [see http://www.nist.gov/srd/reprints.htm (accessed on Oct 30, 2007)].


πrTc2hgrid(Ts - Ta) + πrTc2kTc


2hTc (T - Ta) ) kTcrTc s

πrTc22hgrid(T∞ - Ta) - 2πrTcδkgrid

( )

2hgrid K1(r′Tc) (T - T∞) δkgrid K0(r′Tc) s (A36)

This equation has been written by Guo as (Ts - Ta) ) (Ts - T∞)

-

Figure A2. Geometry of a grid consisting of woven wires.

8hgridδkgrid K1π(r′Tc) (Ta - T∞) - 2hgrid 2 K (r′ ) (Ts - T∞) r 0 Tc

where L represents the length of the plate and dwire represents the diameter of the individual grid wires (which equals the thickness δ of the plate). Ngrid is the number of wires per meter, with a value of

Tc

hgrid +

2hTckTc rTc (A37)

It is also possible to derive an explicit expression for the temperature in the weld, i.e., the temperature measured by the thermocouple

(

(

Ts hgrid + Ta

2hTckTc + rTc

) (

)

8hgridδkgrid K1(r′Tc) ) K0(r′Tc) r 2 Tc

2hTckTc - hgrid + T∞ 2hgrid + rTc

8hgridδkgrid K1(r′Tc) K0(r′Tc) r 2 Tc

)

(A38)

In the theoretical situation that hgrid goes to zero, the equation above shows that the temperature in the weld Ts will equal the temperature of the surrounding gas Ta. If there were no heat loss from the horizontal surface, there would be no dissipation of electric energy and the entire surface would be isothermal. On the other hand, if hgrid is very large, it might be that the temperature in the weld actually becomes higher than in the rest of the grid. This happens because the heat, which may be removed by conduction via the cross-sectional area onto which the thermocouple is welded, is less than what may be removed by radiation and convection from the same area. However, for realistic values of hgrid, this is not the case and the thermocouple will act as heat sink. The value of hgrid will be discussed further in the next section. A2. Thermal Properties of the Grid. In the derivation of eq A37, it was assumed that there is a horizontal surface, which has a heat-transfer coefficient hgrid (comprising convection as well as radiation) between said surface and the gaseous environment. However, the real situation is different; the surface is a heated wire mesh, also known as a heated grid. To evaluate the thermal properties of the grid, it must be realized that it is not a homogeneous body (i.e., a plate) but consists of individual wires, which are woven together, as shown in Figure A2. This difference affects the heat conduction coefficient, the heattransfer coefficients, and the effective surface area, so that it must be corrected. Heat Conduction through the Grid. Because the conduction coefficient is directly linked to the cross-section surface, the model can be corrected by multiplying the conduction coefficient with a correction factor Aeff kgrid ) Aeffkwire

(A39)

In this equation, kwire is the thermal conductivity of the wire material. The correction is achieved by summing up the crosssection areas of all of the wires of the grid and dividing it by the cross-section area of the plate π NgridL dwire2 4 π ) Ngrid dwire Aeff ) Ldwire 4

(A40)

Ngrid )

1 dwire + dap

(A41)

where dap is the aperture width. The sum of the aperture width and the wire diameter equals the mesh width. The thermal conductivity of the grid material AISI 316 L is reported as 16.2 W m-1 K-1 at 100 °C and 21.4 W m-1 K-1 at 500 °C.26 However, it may not rise further with temperature, because the bulk consists of iron, for which the thermal conductivity actually decreases with temperature.27 Therefore, the thermal conductivity was taken to be constant at 20 W m-1 K-1. ConVectiVe Heat Transfer. The convective heat transfer from the grid to the ambient is Qgrid,c ) Seffhwire,c(Tgrid - Ta)

(A42)

In which Seff is the effective surface, which is given for a grid with length L, width W, and thickness of dwire by Seff ) πdwireLWNgrid + πdwireWLNgrid - LWNgrid2dwire2 (A43) The last term in this equation represents the overlap between the wires and may be neglected if the wires are far apart from each other. The derivation of the surface averaged convective heat-transfer coefficient, hwire,c, can be performed using the average Nusselt number, which represents the ratio of the actual heat transferred from the surface of the cylinder to the ambient gas and the occurrence of heat conduction through the ambient gas. The average Nusselt number can be defined by NuD )

hwire,cdwire ka

(A44)

where ka is the heat conduction coefficient of the ambient gas. If the Nusselt number is known and the temperature, pressure, and composition of the ambient gas are given, the heat-transfer coefficient can be calculated. In the current experiment, the ambient gas is nitrogen, kept at a pressure of about 1 atm, and the temperature is assumed to be around 300 K. The Nusselt number, applied to a horizontal cylinder, can be calculated using this information of the ambient gas for natural convection. For this case, an empirical formula is used,25 requiring the Rayleigh and Prandtl numbers

{

NuD ) 0.60 +

0.387RaD1/6 0.559 9/16 8/27 1+ Pr

[ (

) ]

}

2

(A45)

Both the Rayleigh and Prandtl numbers depend upon the properties of the surrounding gas. The Rayleigh number is a combination of the Prandtl and Grashof numbers and can be written as


RaD ) PrGrD ) Pr

Prins et al.

gβ(t - ta)D3

ν2 where β is the coefficient of linear expansion, which is estimated to be 1/ta,28 ν is the kinematic viscosity, and g is the gravitational acceleration. The calculation of the averaged Nusselt number is valid for values of RaD between 10-5 and 1012. The Rayleigh number is also used for calculating the heattransfer coefficients of the thermocouple wires to the environment. The averaged Nusselt number applied to vertical cylinders28 can be calculated from D 0.05 (A47) L which is valid for RaD(D/L) e 0.05, a condition which is easily satisfied for a typical thermocouple diameter D of 0.1-0.2 mm and length L of around 30 mm. RadiatiVe Heat Transfer. The radiant heat transferred from the grid to the ambient is given by29

(

NuD ) 0.93 RaD

)

Qgrid,c ) Seffhwire,r(Tgrid4 - T∞4)

hwire,r )

(A46)

(28) Janna, W. S. Engineering Heat Transfer, 2nd ed.; Van Nostrand Reinhold: London, U.K., 2000.

(A49)

where ε is the emissivity coefficient of the materials of the wires, representing a fraction of the blackbody radiation and σ is the Stefan-Boltzmann constant ()5.6704 × 10-8 W m-2 K-4). The emissivity coefficient of the grid material (AISI316 L stainless steel) may vary but was estimated at 0.2. A3. Thermal Properties of the Thermocouple. For a K-type thermocouple, the wires are made of nickel-chrome or nickel-aluminum, also called chromel and alumel. To obtain a less complicated calculation, only the heat conductivity of chromel is considered kTc ) 0.0191T + 53.22

(A50)

The same approach is followed for the radiant heat transferred from the thermocouple to ambient. The emissivity of a chromel thermocouple wire can be described by30 εTc ) 0.0001T + 0.0595

(A48)

If we wish to write this equation in the same form as eq A43, the radiation heat-transfer coefficient can be expressed in the following way:27

εσ(Tgrid4 - Ta4) (Tgrid - Ta)

(A51)

EF800419W (29) Eckert, E.; Goldstein, R. Measurements in Heat Transfer, 2nd ed.; Hemisphere Publishing: New York, 1976. (30) Sasaki, S.; Masuda, H.; Higano, M.; Hishinuma, N. Int. J. Thermophys. 1994, 15 (3), 547–565.

Visualization of Biomass Pyrolysis and Temperature Imaging in a

Recommend Documents